Should database studies effect patient management and clinical trial design? Discussion of abstracts #4010 and #4011 Howard S. Hochster, MD Professor of Medicine, NYU School of Medicine Director, GI Program NYU Cancer Institute
What kind of database? Prospective clinical data –SEER data base Published clinical trial data –Validated and audited –Accurate for trial result reporting Amalgamated (“pooled”) clinical trial data
Why meta-analysis? Provides greater power than any single trial Answer subset questions that no individual trial could address Findings more robust due to real world heterogeneity Examines the consistency of findings across trials
The Magic of Meta-analysis Σ 95% CI Its significant!!
Wagner, A. D. et al. J Clin Oncol; 24: Fig 4. Effect of fluorouracil (FU)/cisplatin (P)/anthracycline combinations versus FU/anthracycline combinations (without cisplatin) Gastric Cancer: Meta-analysis of adjuvant chemotherapy trials
Pitfalls of meta-analysis Publication bias –‘Positive’ trials more likely to be published rapidly, in more prominent journals –Will skew results if only these trials included Prevent bias by –Thorough literature search –Explicit inclusion/ exclusion criteria –Test for bias: funnel plot
Meta Analysis vs Pooled Analysis Meta analysis: A comprehensive systematic review to address a question using all available data Pooled analysis: Trials are selected for inclusion based on quality, size, relevance –Can re-analyze data set for new and consistent endpoints across studies
Specific Examples of Pooled Analyses Overall treatment effects Subsets –Benefit of chemotherapy in the elderly –Benefit of chemotherapy by stage/grade/# nodes/sex/site –Benefit of chemotherapy by MSI status Consistency across endpoints –Disease Free vs Overall Survival
Is 2-year DFS a “good” surrogate for OS?
CM Total: 43 treatment arms; 20,898 pts 3517QUASAR 867GIVIO 905GERCOR718NSABP C INT NSABP C SWOG FFCD 878N NCIC 915N Siena 2176NSABP C05408N NSABP C04926INT NSABP C03247N NTrialN Active ControlNo Treatment Control ACCENT: Trials included Sargent, 2008
CM Recurrence Rate by 6 mo Intervals Sargent, 2008
CM Time from Recurrence to Death Sargent, 2008 A problem??
Hazard ratios: DFS vs OS OS HR = * DFS HR
ACCENT update: 6 adjuvant trials added TrialAccrual Period # patientsExperimental treatment arm % stage III MOSAIC FOLFOX460 X-ACT Capecitabine100 NSABP C Uracil/tegafur53 NSABP C FLOX71 CALGB IFL100 PETACC FOLFIRI71 Total addition 12,676 patients Median follow-up on living patients 6 years Median survival following recurrence: 20 months
Recurrence Rate over Time 83% of recurrences occur within the first 3 years Same as previous analysis
Within Trial Hazard Ratios for 2 Year DFS vs 5 & 6 Year OS WLS R 2 = 0.58 Copula R 2 = 0.37 WLS R 2 = 0.76 Copula R 2 = 0.49
Forest Plot: 2 Yr DFS v 6 Yr OS Stage III DFS OS
Conclusions: DFS as a stage III endpoint Concordance high for DFS with both 5 & 6 yr OS DFS at 2 and 3 years very similar Strong association DFS yrsOS yrsWLS R 2 Copula R
Good news and bad news: Good news: patients living longer with advanced CRC –5 yr survival 20%; median > 2 yrs Bad news: 5-yr survival follow-up no longer sufficient for adjuvant trials –Maybe we need 7-year follow-up (deGramont, 2008) Good news: 3-yr DFS is predictive for 5-yr OS Better news: 2-yr DFS just as good as 3-yr DFS for predicting 6-yr OS
Do elderly patients (>70 yrs) benefit from newer adjuvant chemotherapy regimens?
Issues Small numbers of elderly patients Under-represented in clinical trials Difficult to perform single adjuvant trial for the elderly & questionable use of resources Prior analyses were well received Unplanned subset analysis –(but pre-specified by ACCENT group)
Questions that might explain these observations in the elderly? Adequate analysis? How does it compare with other analyses? Is there a reason to believe combination therapy affects elderly differentially in the adjuvant therapy as c/w advanced disease? Function of other effects on survival endpoint? –Should we be looking at disease specific OS?
Is this the correct analysis?
ACCENT update: 6 trials added TrialAccrual Period # pts% pts ≥70yrs Experimental treatment arm † % stage III ‡ MOSAIC FOLFOX460 X-ACT Capecitabine100 NSABP C Uracil/tegafur53 NSABP C FLOX71 CALGB IFL98 PETACC FOLFIRI71 † Compared to control arm of intravenous 5-flourouracil (IV 5-FU) and leucovorin (LV) ‡ Remaining patients were stage II or unknown
Forest Plots of Hazard Ratios Disease-Free Survival
Forest Plots of Hazard Ratios Overall Survival
Efficacy – oxaliplatin-based therapy * Values < 1 favor experimental arm Age Endpoint HR (95% CI) Experimental v Control IV 5-FU/LV Deaths within 6 mo Exp v Control % (p-value) DFS * OS * TTR * <70 n = 3, (0.68,0.86) 0.81 (0.71,0.93) 0.76 (0.67,0.86) 0.81 v 0.81 (p=1.0) ≥ 70 n = (0.80,1.35) 1.19 (0.90,1.57) 0.92 (0.69,1.23) 2.57 v 1.37 (p=0.25) Interaction of age by treatment p-value
How do these data compare with other data on elderly and combination chemotherapy for CRC? Is there a reason to expect a differential benefit on older patients receiving adjuvant therapy compared with advanced disease?
Prior age-related analysis Metastatic setting Folprecht JCO N=2,692 (22% 70 yrs) 4 trials of irinotecan-based therapy Improved PFS, trend to improved OS for elderly w/addition of irinotecan Goldberg JCO N=3,742 (16% 70 yrs) 4 trials of oxaliplatin-based therapy Similar survival benefit and toxicity in age subgroups Adjuvant setting Sargent NEJM 2001 – N=3351 (15% 70 yrs) 7 trials of 5-FU + levamisole/leucovorin v surgery No significant interaction observed between age and efficacy of treatment
30 Studies Included 1 Andre et al, NEJM 2004; 2 Goldberg et al, JCO 2004; 3 de Gramont et al, JCO 2000; 4 Rothenberg et al, JCO 2003 Sargent, et. al, Proc ASCO 2006
31 Treatment Benefit Consistent across age groups Consistent across age groups P/DFS: Same hazard ratio P/DFS: Same hazard ratio <70 year olds: HR = 0.7 <70 year olds: HR = 0.7 > 70 year olds HR = 0.65 (p=0.42) > 70 year olds HR = 0.65 (p=0.42) Advanced disease trials Advanced disease trials Response rate: Same odds ratio for response in young vs. old (p=0.38) Response rate: Same odds ratio for response in young vs. old (p=0.38) OS: Same hazard ratio (young HR= 0.77) vs. old HR = 0.82) (p=0.79) OS: Same hazard ratio (young HR= 0.77) vs. old HR = 0.82) (p=0.79) Consistent across age groups Consistent across age groups P/DFS: Same hazard ratio P/DFS: Same hazard ratio <70 year olds: HR = 0.7 <70 year olds: HR = 0.7 > 70 year olds HR = 0.65 (p=0.42) > 70 year olds HR = 0.65 (p=0.42) Advanced disease trials Advanced disease trials Response rate: Same odds ratio for response in young vs. old (p=0.38) Response rate: Same odds ratio for response in young vs. old (p=0.38) OS: Same hazard ratio (young HR= 0.77) vs. old HR = 0.82) (p=0.79) OS: Same hazard ratio (young HR= 0.77) vs. old HR = 0.82) (p=0.79) Sargent, et. al, Proc ASCO 2006
32 Age < 70Age > 70 DFS - Adjuvant Sargent, et. al, Proc ASCO 2006
33 Age < 70Age > 70 PFS - First Line de Gramont, Goldberg Studies Sargent, et. al, Proc ASCO 2006
34 Age < 70Age > 70 OS - First line de Gramont, Goldberg Studies Sargent, et. al, Proc ASCO 2006
A dilemma? How to integrate into patient management? Only “best” elderly are entered in trials Previously benefit was shown –Adjuvant and advanced disease with both FU/LV FOLFOX Final analysis of MOSAIC trial shows small survival differences –Stage III vs. Combined Stage II/III But these two studies seem to suggest a lack of benefit –Adequate power given a 2 study meta-analysis?
Is there chemo “after 70”? Elderly benefit from adjuvant and advanced disease FU/LV (and advanced disease FOLFOX) –Possible differential effects with combination therapy –Possible survival influences of intercurrent disease Toxicities and drug administration may differ in the elderly as new agents are introduced –Especially biologics However, this meta-analysis is based on 2-trial analysis –MOSAIC and NSABP C-06 This observation can inform trial design (especially with newer metrics) –but for individual management? More data are needed: Addition of C-08 and AVANT data may give us a better analysis of elderly patients and adjuvant oxaliplatin
ACCENT Collaborators S Wieand, G Yothers, M O’Connell, N Wolmark – NSABP J Benedetti, C Blanke – SWOG R Labianca – Ospedali Riuniti (Italy) D Haller, P Catalano, A Benson – ECOG C O’Callaghan – NCIC JF Seitz – University of the Mediterranean (France) G Francini – University of Siena (Italy) A de Gramont, T Andre – GERCOR R Goldberg, L Saltz, J Meyerhardt, N Jackson – CALGB M Buyse – IDDI (Belgium) R Gray, D Kerr – QUASAR D Sargent, A Grothey, S Alberts, B Bot, E Green, Q Shi –Mayo Clinic C Twelves -University of Bradford (UK) J Cassidy – University of Glasgow (UK) F Sirzen – Roche ; L Cisar - Pfizer E Van Cutsem –University Hospital Gasthuisberg (Belgium); A Sobrero - Ospedale San Martino (Italy) OUR THANKS!